Trashbytes wrote on 2025-03-21, 12:34:
hrmmm 8000 series card ... smells like bumpgate, have a couple of 8800 cards with similar problems and baking them both allowed them to work again for a while, but baking is not a fix and they will break again.
This. ^ Well, kind of.
The 8800GT is not actually affected by the "bumpgate" issue (anything G9x) isn't - that would be the G8x era of cards and earlier.
But it's moot detail really, since the root cause of the failure is still the same: GPU chip heats too much, causing expansion/contraction of materials to break the solder bumps between the GPU silicon die and GPU substrate.
Trashbytes wrote on 2025-03-21, 12:34:
I would say reflow the GPU die with hot air and see if it improves regardless of the "It aint the solder crowd", right now you cant make it any worse since broken is broken.
Well, it really ain't the solder (I know I already said this in another thread) for sure.
But I'm with you here that it's almost always worth it to try a reflow as you really have nothing to loose 😀 ... that is, unless you plan to send it to someone to do that for you "professionally" (which, TBH, is kind of a waste of time / money for these cards.)
Trashbytes wrote on 2025-03-21, 12:34:
Also give them both a good clean and some fresh paste, it wouldnt hurt, also dont bother with the SLI bridge while testing, its one less thing to cause issues.
+1
zuldan wrote on 2025-03-22, 05:04:
5. Heated the GPU with a hot air station for 3 minutes at 100c (212F) then a further 10 minutes at 200c (392f). I used a temperature probe to make sure the hot air station was actually providing the correct temperature prior to the reflow.
Only 200C, eh?
That's not really enough for the (lead-free) solder to melt. For lead-free solder to melt, you really need to reach at least 220-225C for around 10-20 seconds, or 230C for around 10 seconds. If you look into the datasheets, some SMD components show what a standard reflow profile looks like... and you can achieve that at home on your stove top or with a heat gun, provided you have a temperature probe to monitor temperatures.
So anyways... you didn't really reflow it. More of a re-heat job... which is just as good, IMHO... in most cases anyways. 😀
Now if anything, this further proves what I said in another thread that it's not the solder failing between the GPU substrate and the GPU PCB. It's the solder bumps between the GPU die and substrate that are failing. Re-heating / re-flowing simply causes the GPU die and substrate to expand and contract again, albeit much more due to the higher temperatures than what the GPU can reach under normal operating condition (about 75-85C for these cards under max load, depending on room temperature and case cooling.) At 200C, the greater expansion and contraction between the materials sometimes causes the solder bumps to temporarily make contact again.
So if you want this repair to last, you need to go to a whole new level with the cooling. Otherwise the normal running temps for these GPUs (again, around 75-85C typically) will quickly bring back the original problem again.
zuldan wrote on 2025-03-22, 05:04:
She's working 100%. I've run 3dMark06 multiple times and it's rock solid. Hopefully she'll last with the slow reflow method and good active cooling.
What temperatures are you getting under max load after 10-15 minutes of it running?
At 60-65C, the problem will come back again, albeit a little slower.
At anything above 70C, expect the repair to last no more than a few weeks to a few months... or maybe longer if you don't use the card too often.
But really what you need to do is rip off that shitty stock cooler and put something double-slot on it that really can cope with 100+ Watts of TDP. Or water cooling. Actually, that may be the only thing that will keep these GPUs cool enough while still allowing them to go into a SLI configuration.
Trashbytes wrote on 2025-03-22, 05:51:
Ive seen people mod the side panels to allow 120/140mm fans to be attached to help draw in cool air that way, was quite popular with some case manufactures as well to add fans there .. I have a HAF932 that has a 200mm fan in the side panel for cooling the GPUs. Worked really good for the dual GTX 580s I had in that system!
From experience, I find that adding extra fans to blow fresh air onto the GPUs won't really help to lower temps more than 2-3 degrees C... maybe 5C in a case with poor airflow (card with passive heatsinks excluded, of course.)
On a card that already runs 75-85C under load, 5C is not really that considerable of a temperature drop. You can do better by simply using software like MSI Afterburner to crank up the fans to 100% when the GPU is under load... and for these cards, I believe that can drop the temperature down to maybe 65-70C at best - a good 10C lower than using the stock fan profile. But really, these 8800 GT cards should NEVER have came out with single-slot coolers. A cooler the size that was on the 8800 GTS/GTX or the Radeon HD4870/90 is a lot more inline with what these should have had.
So yeah.... either water cooling or crazy cooler modding is about the only thing that can keep these cards cool enough. 125W TDP is 125W TDP. I mean, would you ever put that small GPU cooler on top of a 125W TDP CPU and say "I'm sure this will run fine" - probably not. Saying that, I can only think of that Sonic meme with the flames roasting in the background and Sonic saying "this is fine". 🤣
The Serpent Rider wrote on 2025-03-23, 05:42:
Overclocking should be fine. You can't bump voltage on 8800 cards and raising GPU clock about 20-25% won't increase temperatures in any meaningful way.
I still wouldn't.
But if you do want to play that route (messing with GPU clocks and voltages), then may I suggest something else - how about mod the BIOS to lower the GPU voltage and clocks? This is what the 9800 GT "green edition" (GE)/ "power-efficient" (PE) / "energy efficient" (EE) / "ECO" editions do. FWIW, these are based around the same G92 cores IIRC, but at slightly lower core clocks and also lower core voltages... resulting in a TDP of 70-ish Watts, IIRC... which is a considerable drop from the original 125W TDP.
That and good cooling *may* just extend the "reflow" card's life long enough that it serves you well for many years... or it may not.
When it comes to reflows, "luck" will also play a big part in how long a "fix" lasts. Of course, with good cooling, chances are you will well be extending the life of however long that fix would have lasted if you didn't do it otherwise. So definitely worth working on improving the cooler and certainly NO OC'ing. (unless you don't give a crap about the card - if that's the case, F--- it and OC to the moon 🤣 )
